Fix duplicate logs & backlogs on container restart

[ebr]

Connects to #187

• Nothing needs to be done on "restart" event, because a "start" event always(?) follows it. A pumpLogs was being called on both restart and start, which caused duplicate entries after a restart.
• Add support for a BACKLOGS environment variable:
  • Unless specified BACKLOGS=false, it will default to true for all other values, consistent with current behaviour;
  • BACKLOGS=false is equal to docker logs --tail=0

[olivergrahl]

+1 for getting this merged. We start containers using cron every couple of minutes, so we get all the log history sent again on every single start.

[josegonzalez]

We're instead looking at #209 as a more elegant solution to the problem. Thanks for the +1 though.

[olivergrahl]

Any news on this? It's really an issue for us, as we start containers every couple of minutes, and monitor the logs via Papertrail alerts. Once a string that would trigger an alert gets logged from the container, it will be logged again and again on each container start, resulting in the alert being triggered again and again... Any idea how we can prevent that? Thx!

[ebr]

@josegonzalez are we still looking at #209? Shall I rebase this PR, or abandon it?
I'm ok with either, as long as we get a fix for this issue, since it's blocking us from running master in hopes to fix or alleviate #216

[ebr]

@josegonzalez (cc @olivergrahl) this is rebased, tested in my Staging and Prod and looking good. Takes care of the backlog dump with BACKLOG=false. double pump was already fixed elsewhere.

[michaelshobbs]

merged in #280